Daily Physical Activity, Sedentary Behavior and Alcohol Use in At-Risk College Students

Abstract

Background

The college years present an opportunity to establish health behavior patterns that can track across adulthood. Health behaviors tend to cluster synergistically however, physical activity and alcohol have shown a positive association.

Purpose

This study applied a multi-method approach to estimate between- and within-person associations between daily physical activity, sedentary behavior and alcohol use among polysubstance-using college students.

Methods

Participants were screened for recent binge drinking and either tobacco or cannabis use. They wore an activPAL4 activity monitor and a Secure Continuous Remote Alcohol Monitor continuously in the field for 11 days, and completed daily online questionnaires at the beginning of each day to report previous day physical activity, sedentary behavior, and alcohol consumption.

Results

Participants (N = 58, M_age = 20.5 years, 59% women, 69% White) reported meeting national aerobic physical activity guidelines (75%) and drinking 2–4 times in the past month (72%). On days when participants reported an hour more than usual of daily sedentary behavior, they reported drinking for less time than usual (γ = −.06). On days when participants took 1,000 more steps than usual, the longest episode of continuous transdermal alcohol detection was shorter (γ = −.03).

Conclusions

Daily physical activity and sedentary behavior were negatively associated with time-based measures of alcohol use with the lowest risk on days characterized by both activity and sedentary behavior. Intensive longitudinal monitoring of time-based processes can provide new insights into risk in multiple behavior change and should be prioritized for future work.

Physical activity, alcohol use, sedentary behavior, polysubstance use, college students

The college years present opportunities for establishing health habits that can track across adulthood. Movement-related behaviors (i.e., physical activity and sedentary behavior) and substance use typically change in unhealthy ways during this period [1, 2]. To the extent that these health behaviors cluster, interventions to improve either behavior may have consequences for multiple behaviors. Physical activity and alcohol use have consistently clustered such that increased physical activity has been associated with increased alcohol use but those findings have been based largely on cross-sectional, self-report data [3]. The availability of sensors to monitor health behaviors has opened new possibilities for tracking these behaviors in the natural context of daily life to enrich understanding of their associations.

Key Health Behaviors Among College Students

Physical inactivity, sedentary behavior and alcohol use are key health behaviors that can worsen and increase risk during the college years.

College students engage in greater physical activity levels than their non-college peers; however, little more than half (57%) of on-campus students report meeting the physical activity guidelines [4]. Sedentary behavior is the other waking component of the 24-hour activity cycle [5]. Sedentary behavior refers to any waking behavior characterized by low energy expenditure (≤1.5 metabolic equivalents [METs]), while in a seated, reclined, supine, or prone posture [6]. College students are at particular risk for excessive sedentary behavior as a result of their unique lifestyle demands (e.g., sitting in class, studying). Ecological momentary assessment (EMA) studies suggest college students spend as much as 67% of waking time or 11 hours/day engaged in sedentary behavior [7].

Alcohol use is the most common form of substance use among college students. Results from the 2018 Monitoring the Future Survey showed that both past-year and past 30-day alcohol use were higher among college students than their non-college peers (75% vs. 70% and 60% vs. 50%, respectively) [2]. Risky drinking patterns, such as binge drinking, have been flagged as a consistent problem on college campuses. Binge drinking is defined as consuming 4 or more drinks for women and 5 or more drinks for men in one sitting (~2 hours) [8]. Nearly a third (29%) of college students reported binge drinking in the past two weeks compared to 25% of their non-college peers [2]. Unlike insufficient physical activity or excessive sedentary behavior, binge drinking can have adverse consequences for both the user and others around them.

Physical Activity and Alcohol Use

Health behaviors often cluster synergistically such that people who engage in one healthy behavior are more likely to engage in other health behaviors (e.g., exercising and eating well) [9]. In contrast, physical activity is associated with greater alcohol use [3, 10, 11]. This antagonistic clustering is unlike associations between physical activity and other substance use behaviors where more active people are less likely to use tobacco or cannabis [12–15]. The positive association between physical activity and alcohol use has been identified at both the between- and within-person levels [16–18]. In other words, people who engage in more physical activity consume more alcohol in general (between-person association) and on days when people engage in more physical activity than usual, they consume more alcohol than usual (within-person association). This antagonistic clustering of physical activity and alcohol use, including binge drinking, has been identified among college students [3].

Three methodological limitations complicate interpretations of the available literature. First, the existing literature has largely been based on self-reports of these behaviors [3]. When device-based measures have been used with college students, it has been limited to devices for measuring physical activity [19]. Self-reports of alcohol use often only reflect the intensity of alcohol use (i.e., number of standard servings of alcohol consumed) and reveal little about the topography or patterning of daily alcohol use [20]. Measures such as drinking duration or length of time that alcohol is detected transdermally may provide important context for understanding associations between physical activity and alcohol use. These time-based measures of alcohol use could represent the length of risk rather than relying solely on the intensity of risk. Understanding these nuances in risk could help explain the alcohol’s association with physical activity and enrich understanding of the risk conferred by this association. To date, no study has used sensors to measure both physical activity and alcohol use in real-time. The present study extends the literature on physical activity and alcohol use by using an ecological momentary assessment (EMA) research design that integrates self-reports and device-based measures of both behaviors in daily life.

Second, compared to the literature on physical activity and alcohol use, less is known about the association between sedentary behavior and alcohol use. Among adolescents, sedentary behavior was associated with the increased alcohol consumption [3]. The literature on college students is too sparse to draw conclusions [3]. It is possible that the association between sedentary behavior and alcohol use among college students is positive like it is for adolescents. Many studies operationally defined sedentary time as screen time and little is known about other domains of sedentary behavior (e.g., transportation, at work or school, leisure time outside of TV time) that contribute to total sitting time [3]. The present study will advance the literature by capturing both device-based measures of sedentary time and self-reports of domain-specific sedentary time to estimate associations with alcohol use.

Finally, previous studies with college students have not differentiated between single- and polysubstance use patterns. Polysubstance use refers to a pattern of using more than one drug during a specified timeframe (e.g., past 30 days) and includes both those who use multiple substances concurrently and those who co-use within the same occasion [21]. Polysubstance use is associated with additional adverse health and social outcomes when compared with single substance use [22, 23]. Polysubstance use has been highlighted recently due to a new pattern of simultaneously using alcohol and cannabis that is growing in popularity among college students [24]. Polysubstance users may have different motivations for engaging in substance use compared to single-substance users [25], and these differing motivations may influence associations with other health behaviors (i.e., physical activity, sedentary behavior). Only two studies have examined physical activity and polysubstance use, and both were limited to cross-sectional data [26, 27]. The present study will extend this literature by screening for polysubstance users who engage in risky drinking patterns to estimate between- and within-person associations in this high-risk sample of college students.

Purpose

The Physical Activity, Recreational THC and Ethanol Use (PARTE) study was designed to estimate between- and within-person associations between physical activity, sedentary behavior and alcohol use among a sample of polysubstance using college students who engage in risky drinking patterns. This study was designed to add rigor to the literature by incorporating both self-report and device-based measures of each behavior, simultaneously modeling physical activity and sedentary behavior in each model, and delimiting the sample to high-risk polysubstance users. An 11-day EMA protocol was designed to capture two social weekends when alcohol use was expected to be greatest in this population. Key hypotheses were that (i) people who engage in more physical activity would consume more alcohol, (ii) on days when people engaged in more physical activity than usual, they would consume more alcohol than usual, (iii) people who engaged in more sedentary behavior would consume more alcohol, and (iv) on days when people engaged in more sedentary behavior than usual, they would consume more alcohol than usual.

Methods

Participants

Participants were recruited from a population of male and female college students at a large university in the Northeast who were high-risk alcohol users with a polysubstance use history. The study was advertised as research examining physical activity and alcohol use in the context of daily life. Participants were recruited via fliers on campus and in local downtown establishments, departmental list serves, and an online institutional portal for research recruitment. Inclusion criteria included being age 18–25 years, enrollment at a university at the time of the study, binge drinking at least twice in the past month, and reported use of either tobacco or cannabis at least once in the past year. Individuals who reported a history of substance dependence were excluded because they represent a specialized segment of the population that likely need to be studied separately from the general population.

Measures

Demographics

At baseline, participants reported sex, age, racial and ethnic background, student status, employment status, Greek life status and athlete status. A research assistant measured the participants’ height and weight in duplicate.

Baseline polysubstance use

Polysubstance use data were collected to characterize the sample. At baseline, participants reported past 30-day frequency of cigarette use (options: 0 to 30 days; “less than 1 cigarette/day” to “26+ cigarettes/day”) as well as lifetime quantity of cigarettes used (options: “Never” to “5 or more packs”). Participants also reported lifetime and past 30-day frequency of marijuana use (options: “0 occasions” to “40+ occasions”).

Baseline alcohol use

Participants completed the 10-Alcohol Use Disorders Identification Test (AUDIT). Scores from the AUDIT are included to describe the participant sample. Scores ranging from 8 to 12 indicate harmful or hazardous drinking. Scores of 13 or more among women and 15 or more in men are considered an indication of potential alcohol dependence [28]. The AUDIT has both high sensitivity (.76) and specificity (.79) [29].

Baseline physical activity and sedentary behavior

At baseline, self-reported physical activity was assessed using the International Physical Activity Questionnaire – Short version (IPAQ) [30]. Participants reported on frequency and duration of light, moderate and vigorous activity in the past 7 days. Using standard IPAQ scoring protocol, the total physical activity volume score was calculated to represent the total estimated energy expenditure during physical activity for that week [31]. The IPAQ has been shown to produce repeatable data and has a criterion validity comparable to other self-report measures of activity (r = 0.30) [30].

At baseline, self-reported sedentary behavior was assessed using the Marshall Sitting Questionnaire [32]. Participants reported how many hours or minutes they spent sitting per day in five common domains of sedentary behavior (e.g., while traveling, while at school or work, watching television). Participants were asked to report sitting time for both a typical weekday and a typical weekend day. Time spent sitting across domains was summed to provide total scores for daily sitting time on weekdays and weekends. Reliability and validity range for this questionnaire but are highest for weekday reports, specifically in the domains of sitting at work and using a computer at home (r = .69–.84) [32].

Daily alcohol use

At the beginning of each day, participants self-reported the number of standard servings of beer (12 oz), wine (5 oz), and liquor (1.5 oz) they consumed on the previous day. A graphic depicting standard serving sizes for beer, wine and liquor was provided at the beginning of the questionnaire. Responses were used to provide a daily drinking intensity score. Participants also reported the times they started and stopped drinking. The interval between these times was calculated to provide a daily drinking duration score.

The Secure Continuous Remote Alcohol Monitor (SCRAM; SCRAM CAM, Highlands Ranch, CO) device is the current gold standard for ambulatory assessment of alcohol consumption [33]. The SCRAM device was worn around the ankle for 11 days to provide estimates of transdermal alcohol content (TAC) every 30 minutes [33]. SCRAM collects timestamped data on alcohol exposure, by sampling insensible perspiration above a small surface area around the ankle. This reading is defined as transdermal alcohol concentration (TAC). IR voltage and device temperature sensors are also included and used to verify device wear. Specificity of the SCRAM device is high (~88%) [34]. This study followed the methods of Roache and colleagues [35] for improving the sensitivity of episode detection. The SCRAM device can reveal a more detailed topography of daily alcohol use by indicating both the intensity and timing of total alcohol concentrations. The SCRAM device does not provide any real-time feedback. In our study, data were stored on the device and only viewable by the researcher once the device had been returned.

Both self-report and device-based data were used to characterize daily drinking. Drinking intensity was quantified as daily self-reported total standard servings of alcohol and daily peak transdermal alcohol concentration (TAC) recordings from the SCRAM device. Peak TAC is highly correlated with blood alcohol concentrations (particularly at higher levels of alcohol consumption) [36]. It can represent either greater consumption, faster consumption, or both [34]. Drinking episodes were defined as a consecutive string of non-zero TAC values with no more than one zero TAC value in between non-zero values [35]. Three variables were calculated to characterize the timing of daily alcohol consumption. Drinking duration was calculated as the difference between self-reported start and stop times for drinking each day. Time to peak drinking intensity was calculated using data from the SCRAM device. Once daily peak TAC was identified, the time stamp for the beginning of that drinking episode was recorded. Time to peak drinking intensity was calculated as the difference between the timestamp at which peak TAC was reached and the timestamp for the beginning of that drinking episode. Longest duration of transdermal alcohol detection was also calculated using data from the SCRAM device. The longest single drinking episode was defined as the longest string of non-zero TAC values for each person on each study day. Longest duration of transdermal alcohol detection was calculated as the difference between the timing of the first and last readings for the longest single drinking episode for that day. Although drinking days could have more than a single drinking episode, the longest drinking episode was of interest because of the heightened health risk of extended periods under the influence of alcohol. For example, health and social consequences related to binge drinking increase as duration increases [37].

Daily physical activity and sedentary behavior

At the beginning of every day, self-reported physical activity was assessed using a daily adaptation of the International Physical Activity Questionnaire – Short version (IPAQ) [30, 38]. Participants reported on the duration of intensity-specific physical activity (walking, moderate and vigorous physical activity) from the prior day. The total physical activity volume score was calculated using established IPAQ weights to estimate the total estimated energy expenditure (MET-min) for each day.

At the beginning of every day, self-reported sedentary behavior was assessed using a daily adaptation of the Marshall Sitting Questionnaire [32]. Participants reported how many hours or minutes they spent sitting the day before in five domains (e.g., while traveling, while at school or work, watching television). Time spent sitting for each of the domains was summed to provide a total duration for daily sitting time.

ActivPAL4 monitors were used to assess momentary physical activity and sedentary behaviors (PAL Technologies Ltd, Glasgow, Scotland). This device senses acceleration and classifies physical activity and sedentary behavior in real-world contexts [39]. The activPAL device is wrapped in a nitrile sleeve to waterproof the device, and then is attached to the upper anterior midline of the thigh using hypoallergenic Hypafix medical tape [40]. The primary measure of physical activity volume was step counts [41]. Previous literature examining associations between physical activity and alcohol use have defined physical activity as both total volume and bouts of activity. Step counts were chosen to provide a more normally distributed picture of total physical activity volume across the day. The primary measure of sedentary behavior was the duration of waking sedentary time (i.e. sitting time). ActivPAL devices have shown to be reliable for accurately reporting step counts in the field and classifying physical activity and sedentary behaviors [39, 42]. Daily data were considered valid if the participant had at least 10 hours of waking wear time during the 24-hour period. Participant sleep log data were used to verify wake time for each day [40]. Participants were asked to wear the device for at least 20 out of 24 hours each day on at least 9 out of the 11 days of the study.

Participants completed a daily wear time log detailing the time they woke up and fell asleep as well the time(s) of any device removal(s).

Procedures

Prospective participants were screened for eligibility in telephone interviews (see Appendix for details). Eligible participants visited the lab on a Wednesday (Study Day 0) to provide informed consent, complete baseline questionnaires using Qualtrics on a lab computer, and receive training on all study procedures including how to wear and remove the activPAL4 and SCRAM devices. The researcher fit the participant with a waterproofed activPAL4 activity monitor and a SCRAM device to wear continuously over the next 11 days (except while bathing or swimming). Participants also received a paper logbook in which to record sleep and device non-wear times. Finally, participant emails were verified to ensure that the daily questionnaires would be successfully delivered.

The ambulatory monitoring period began the next day (i.e., Thursday) and continued for 11 complete days. This data collection period was selected to include two weekends to coincide with expected alcohol use in this population [43, 44]. During these 11 study days participants wore the activPAL4 and SCRAM devices continuously and completed daily online questionnaires at the beginning of each day. The beginning-of-day questionnaires included reports of past-day physical activity, sedentary behavior and substance use. These self-reports were completed via Qualtrics links that were emailed to participants each morning at 9 am. The daily questionnaires took approximately 15 minutes to complete and participants had access to the link for about 12 hours. Reminders to complete these surveys were not sent to participants. Participants were asked to email the PI immediately if there were any issues receiving the emailed link.

Participants returned to the lab on the second Monday (Study Day 12) to return equipment and complete web questionnaires. The researcher debriefed, thanked, and compensated participants for their participation. Participants received $40 cash for completing all procedures in the study with a $10 cash bonus if they completed at least 9 of the 11 daily questionnaires and wore both the activPAL and SCRAM device on 9 of the 11 days for 20 of 24 hours (i.e. 80% compliance).

Data Processing

Pre-processing activity data

Self-reports of total physical activity volume and total sitting time and sensor data from the activPAL were used to calculate daily activity and sedentary scores for each participant. For the self-report data all reports were aligned with the day of use as opposed to day of reporting to create a daily score. The sensor data were collected intensively and then reduced to daily summary measures.

Non-wear time was screened out and days with too much non-wear time (i.e., <10 hours of waking wear time) were considered invalid days. Activity data were annotated to indicate times when paper logs showed that participants were sleeping or awake. When the paper log was missing, activity data were annotated by labeling an extended period of time without step counts as sleep and periods where there was a sedentary-to-upright transition followed by step counts as awake. To identify the transition from awake-to-sleeping and sleep time, the time stamp where the activity file shows an upright-to-sedentary transition followed by an extended period without step counts was labeled as the transition and sleeping. During this sleep period if a few step counts were registered it was considered a bathroom break and was annotated under sleep time unless the step counts were registered for 10+ consecutive minutes. To identify the transition from sleeping-to-awake and wake time, the time stamp where the activity file shows a sedentary-to-upright transition was located. If this transition was preceded by an extended period of time without step counts and followed by a period of regular step counts, then this was noted as the transition and wake time.

Segmenting study days

Days were segmented at 4 am (as opposed to 12 am) primarily to capture alcohol consumption that could occur late at night and extend past midnight. The decision to segment at 4 am corresponded with the activPAL data and sleep logs. The mean step count at 4 am was negligible (M = 0.18, mode = 0) and the sleep logs showed no reports of waking up at 4 am, whereas 11 participants reported waking up between 5 and 6 am with two of those participants waking up at 5 am.

Step counts and waking sedentary duration were aggregated for each valid day. Physical activity and waking sedentary scores were person-mean-centered by redistributing daily score variance into a person-level mean score (between-person variable) and a daily deviation from that mean (within-person variable) per standard multilevel modeling practices [45, 46].

Pre-processing alcohol data

Self-reported total standard servings of alcohol consumed as well as TAC data from the SCRAM devices was used to calculate daily alcohol use scores. For the self-report alcohol data all reports were aligned with the day of use as opposed to day of reporting to create a daily score. TAC data from the SCRAM device provided drinking intensity and timing of daily alcohol consumption. TAC data from the SCRAM device was taken every 30 minutes. The 48 readings from each 24-hour period were aggregated to create daily scores for drinking intensity and timing of daily alcohol consumption. The TAC data was cleaned, following procedures outlined by Roache and colleagues [35], to remove trivial readings and readings likely to be caused by contamination (i.e., high sudden spikes that immediately declined). High-frequency reading (i.e. readings less than 29 minutes apart) were removed. Readings were also removed if the reading was a single non-zero TAC value bracketed by zero values. Finally, a reading was removed if the first TAC positive value was above 0.15.

Data Analyses

Daily data were nested within people. To accommodate dependencies between observations, multilevel models (MLM) were estimated using SAS software version 9.4 to test relations between physical activity, sedentary behavior and alcohol use (SAS Institute, Cary, NC, USA). First, empty (random intercept only) models were estimated to conduct variance decomposition analyses and estimate the proportion of total variation in each variable attributable to between-person differences. Next, the fit of Gaussian, zero-inflated Poisson, and negative binomial error distributions were compared in an empty (unconditional) model to determine the appropriate model for subsequent hypothesis tests with data that may be zero-inflated.

Five MLMs were estimated to evaluate associations between physical activity, sedentary behavior, and alcohol use in this study. The first two models regressed drinking intensity on physical activity and sedentary behavior. In Model 1, self-reported drinking intensity (total daily standard servings of alcohol) was regressed on self-reported physical activity and sedentary behavior. Self-reported physical activity volume was represented as total estimated energy expenditure from daily light, moderate, and vigorous-intensity physical activity (MET-min). Self-reported sedentary behavior was represented by the total duration of daily sitting time. In Model 2, device-measured drinking intensity (defined as peak daily TAC) was regressed on physical activity and sedentary behavior. Device-measured physical activity volume was represented as daily step counts from the activPAL monitor. Device-measured sedentary behavior was represented by the duration of seated and reclined postures during waking hours from the activPAL monitor.

The next three models regressed different measures of the timing of daily alcohol consumption on physical activity and sedentary behavior. In these models, physical activity volume and sedentary behavior were defined as above. In Model 3, self-reported drinking duration was regressed on self-reported physical activity and sedentary behavior. In Model 4, device-measured time-to-peak drinking intensity was regressed on device-measured physical activity and sedentary behavior. In Model 5, device-measured longest duration of transdermal alcohol detection was regressed on device-measured physical activity and sedentary behavior.

Fixed effects for physical activity and sedentary behavior were evaluated to test hypotheses. Random effects were estimated for the intercept and slopes of physical activity and sedentary behavior. Covariates in each model included sex, baseline BMI and day of the week (i.e. weekday [Sunday, Monday, Tuesday, and Wednesday coded as 0] vs. social weekend [Thursday, Friday, and Saturday coded as 1]). Drinking patterns, particularly among college students, suggest that Sunday is best classified as a weekday, whereas Thursday typically follows weekend drinking patterns [47, 48]. Alcohol consumption can vary widely between the weekday and weekend, as identified above. Further, men are more likely to both consume alcohol and consume alcohol at higher frequencies [49].

The level-1 model represented the daily repeated measures within each participant, and the level-2 model represented differences between participants. The generic model that was tested is as follows:

Level-1:	${\mathrm{Y}}_{ti}={\beta }_{0i}+{\beta }_{1i}\mathrm{P}\mathrm{h}\mathrm{y}\mathrm{s}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{A}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{i}\mathrm{t}{\mathrm{y}}_{ti}+{\beta }_{2i}\mathrm{S}\mathrm{e}\mathrm{d}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{a}\mathrm{r}\mathrm{y}\mathrm{B}\mathrm{e}\mathrm{h}\mathrm{a}\mathrm{v}\mathrm{i}\mathrm{o}{\mathrm{r}}_{ti}+{\beta }_{3i}\mathrm{S}\mathrm{o}\mathrm{c}\mathrm{i}\mathrm{a}\mathrm{l}\mathrm{W}\mathrm{e}\mathrm{e}\mathrm{k}\mathrm{e}\mathrm{n}{\mathrm{d}}_{ti}+{\mathrm{e}}_{ti}$	(1)
Level-2:	${\beta }_{0i}={\gamma }_{00}+{\gamma }_{01}\mathrm{U}\mathrm{s}\mathrm{u}\mathrm{a}\mathrm{l}\mathrm{P}\mathrm{h}\mathrm{y}\mathrm{s}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{A}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{i}\mathrm{t}{\mathrm{y}}_i+{\gamma }_{02}\mathrm{U}\mathrm{s}\mathrm{u}\mathrm{a}\mathrm{l}\mathrm{S}\mathrm{e}\mathrm{d}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{a}\mathrm{r}\mathrm{y}\mathrm{B}\mathrm{e}\mathrm{h}\mathrm{a}\mathrm{v}\mathrm{i}\mathrm{o}{\mathrm{r}}_i+{\gamma }_{03}\mathrm{S}\mathrm{e}{\mathrm{x}}_i+{\gamma }_{04}\mathrm{B}\mathrm{M}{\mathrm{I}}_i+{\mu }_{0j}$	(2)
	${\beta }_{1i}={\gamma }_{10}+{\mu }_{1j}$	(3)
	${\beta }_{2i}={\gamma }_{20}+{\mu }_{2j}$	(4)
	${\beta }_{3i}={\gamma }_{30}+{\mu }_{3j}$	(5)

In the level-1 model, Y_ti represented the continuous alcohol use on day t for participant i; β_0i represented the average level of daily alcohol use for participant i; β_1i and β_2i represented the within-person association between alcohol use and physical activity or sedentary behavior, respectively; β_3i represented the within-person association between alcohol use and whether the day is considered a social weekend day (coded as 1) or a weekday (coded as 0); and e_ti represented the usual residual error term. In the level-2 models, γ₀₀ represented the daily alcohol use grand mean for the entire sample; γ₀₁, γ₀₂, γ₀₃ and γ₀₄ represented the between-person association between alcohol use and usual physical activity, usual sedentary behavior, sex or BMI, respectively; µ _0j represented the residual error term. γ₁₀, γ₂₀ and γ₃₀ represented the fixed effects for physical activity, sedentary behavior, and social weekend; and µ_1j, µ_2j and µ_3j represented the deviation of the slopes for physical activity, sedentary behavior, and social weekend from the overall slopes. Given the non-normality of the alcohol data (i.e. excess zeros), multilevel over-dispersed Poisson models were run. This technique models the number of drinks according to a Poisson distribution, while adjusting for overdispersion that results from a large number of zeros. This approach has been used in previous studies examining alcohol use [50]. Multilevel over-dispersed Poisson models handle missing data using a residual pseudo-likelihood estimation, equivalent to a restricted maximum likelihood estimation.

Results

Participants (N = 58, M_age = 20.5) enrolled between September 2019 and March 2020. Data collection was halted when the World Health Organization declared the COVID-19 pandemic and the university paused all face-to-face research activities. Table 1 summarizes participant characteristics. Most participants were women (n = 34, 59%), White (n = 40, 69%) and undergraduate students (n =56, 97%). A quarter of participants (n = 15, 26%) reported participating in Greek life, whereas less than a quarter (n = 10, 17%) reported participating in collegiate, club or IM sports.

Table 1.

Demographic characteristics of the sample (N = 58)

	n (%)	M (SD)
Age		20.5 (1.4)
Women	34 (58.6%)
Caucasian	40 (69%)
Black or African American	8 (13.8%)
Asian	4 (6.9%)
Two or More Races	2 (3.4%)
Hispanic	13 (22.4%)
BMI		24.5 (4.8)
Underweight	2 (3.4%)
Normal Weight	32 (55.2%)
Overweight	18 (31%)
Obese	6 (10.3%)
Student Characteristics
Undergraduate Student	56 (96.6%)
Graduate Student	2 (3.4%)
Participate in Greek Life	15 (25.9%)
Participate in Collegiate, Club or IM Sports	10 (17.2%)

The majority of the participants (98%) completed the entire 11-day study. One participant dropped out after completing 5 days due to the SCRAM device being “too uncomfortable” to wear. One participant’s SCRAM device’s air pump malfunctioned, but 5 out of the 11 days of data were recovered. Available data from all participants were included in all analyses. Out of the possible 638 study days, the majority of those days were valid for the daily questionnaires (n = 505, 79%) activPAL (n = 523, 82%) and SCRAM device (n = 626, 98%).

Baseline Behavioral Characteristics

Table 2 summarizes participant baseline physical activity and sedentary behavior. Most participants (75%) met national guidelines for aerobic physical activity based on their reports of past-week physical activity. Participants reported walking as their most common intensity of physical activity, followed by vigorous activity. Participants reported being highly sedentary on both weekdays and weekend days.

Table 2.

Descriptive statistics for baseline physical activity, sedentary behavior and substance use (N = 58)

	n (%)	M (SD)	Possible range	Observed range
Self-Reported Past Week Physical Activity
Meeting Guidelines	39 (75%)
PA Volume – Walking (MET·mins)		3,196.7 (1,075.6)	0–4,158	693–4,158
PA Volume – Moderate (MET·mins)		1,046.7 (1,270.1)	0–5,040	0-5,040
PA Volume – Vigorous (MET·mins)		1,436.3 (1,376.1)	0–10,080	0-4,800
PA Volume – Total (MET·minutes)		5,781.1 (2,452.5)	0–19,278	2,439–13,998
Self-Reported Past Week Sedentary Behavior
Single Weekday (minutes)		790.5 (172.5)	0–960	450–960
Single Weekend Day (minutes)		754.6 (191.3)	0–960	290–960
Self-Reported Alcohol Use
Drinking Frequency
Monthly or less	14 (24.6%)
2–4 times a month	41 (71.9%)
2–3 times a week	2 (3.5%)
Drinking Intensity on a Typical Drinking Day
1 or 2	22 (38.6%)
3 or 4	21 (36.8)
5 or 6	9 (15.8%)
7 to 9	5 (8.8%)
Baseline AUDIT Score		10 (4.0)	0–40	1–21
No risk	18 (31%)
Hazardous drinking	29 (50%)
Potential Dependence	11 (19%)
Self-Reported Polysubstance Use
Past 30-Day Cigarette Use Frequency
0 days	24 (64.9%)
1 or 2 days	11 (29.7%)
6 to 9 days	1 (2.7%)
10 to 19 days	1 (2.7%)
Past 30-Day Cannabis Use Frequency
0	14 (24.6%)
1–2	12 (21.1%)
3–5	6 (10.5%)
6–9	6 (10.5%)
10–19	7 (12.3%)
20–39	11 (19.3%)
40+	1 (1.8%)

As seen in Table 2, baseline AUDIT scores categorized 50% of the participants as participating in hazardous drinking and 19% as having potential dependency characteristics. Most participants (72%) reported drinking 2–4 times a month and 75% reported drinking 1–4 drinks on a typical drinking day. All participants reported being polysubstance users (i.e. using either tobacco or cannabis) at least once in the past year. Past-month cannabis use at baseline (76%) was more common than past-month tobacco use at baseline (35%).

Daily Behavioral Characteristics

Table 3 describes daily physical activity, sedentary behavior and alcohol use. When reporting descriptive statistics, the weekend was defined differently. For physical activity and sedentary behavior, weekends were defined as Saturday and Sunday and weekdays were defined as Monday through Friday. For alcohol use data, the social weekend was defined as Thursdays through Saturday and the weekdays were defined as Sunday through Wednesday. Across both self-reports and the activity monitor, participants were more active on weekdays than weekends. Participants were also more sedentary during weekdays compared to weekend days. Patterns were similar for self-reports and device-based measures.

Table 3.

Descriptive statistics for daily physical activity, sedentary behavior and alcohol use

	Weekday			Weekend
	M	SD	Range	M	SD	Range
Physical Activity
Self-reported Light (min/day)	102.62	55.9	0–180	101.41	59.2	0–180
Self-reported Moderate (min/day)	44.02	53.7	0–180	40.03	57.5	0–180
Self-reported Vigorous (min/day)	24.95	42.4	0–180	12.9	32.9	0–180
Self-reported Active Time (min/day)	170.64	102.1	0–540	154.06	103.6	0–540
Device-Measured (steps/day)	11,449.8	5,046.7	612–30,458	9,558.6	5,213.2	600–30,298
Sedentary Behavior
Self-Reported (min/day)	537.6	185.0	125–960	508.45	213.0	60–960
Device-Measured (min/day)	448.2	161.6	1.4–888.6	358.9	154.8	19.9–731.5
Alcohol Use
Self-reported Drinking Intensity (standard servings/day)	0.47	1.3	0–10	3.1	4.0	0–15
Self-reported Drinking Duration (min/day)	103.78	79.7	15–300	268.06	185.2	10–950
Device-Measured Drinking Intensity (g/dl)	0.02	0.05	0–0.342	0.07	0.11	0–0.437
Device-Measured Time-to-Peak TAC (min)	68.20	92.19	0–431	140.28	190.37	0-1,260
Device-Measured Longest Duration of Transdermal Alcohol Detection (min/day)	286.42	251.72	0-1,415	358.92	281.82	0-1,429

^aFor physical activity and sedentary behavior, weekends were defined as Saturday and Sunday and weekdays were defined as Monday through Friday.

^bFor alcohol use data, the social weekend was defined as Thursdays through Saturday and the weekdays were defined as Sunday through Wednesday.

Intraclass correlation coefficients (ICC) indicated that only 5–21% of the variability in daily alcohol use was attributable to between-person variance: self-reported drinking intensity ICC = .12, self-reported drinking duration ICC = .06, device-monitored drinking intensity ICC = .21, device-monitored time to peak TAC ICC = .05, and device-monitored longest duration of transdermal alcohol detection ICC = .09. Table 4 presents between- and within-person correlations between self-reported and device-based measures of alcohol use. At the between-person level, people who reported more physical activity recorded more steps, but people who reported more sedentary behavior did not accumulate more sedentary time. The alcohol use data yielded moderate-to-strong positive associations across measurement methods. The within-person correlations are reported for descriptive purposes but cannot be tested for statistical significance because each person contributed multiple data points (i.e., observations were not independent). Based on the magnitude of between- and within-person associations, these two measurement methods shared some overlap but were not redundant.

Table 4.

Pearson correlations between daily self-report and device-based measures

	Self-reported Total MET-mins	Self-reported Sedentary Behavior (Hrs)	Self-reported Drinking Intensity (standard servings/day)	Self-reported Drinking Duration (min/day)
Between-person
Device-Measured Daily Steps	.491**	−.408**	.028	−.091*
Device-Measured Sedentary Behavior (Hrs)	−.110*	−.142**	−.249**	−.203**
Device-Measured Drinking Intensity (g/dl)	.080	.020	.482**	.485**
Device-Measured Time-to-Peak TAC (min)	.133**	.075	.499**	.544**
Device-Measured Longest Duration of Transdermal Alcohol (min)	.009	.069	.633**	.615**
Within-person
Device-Measured Daily Steps	.420	−.153	.154	.169
Device-Measured Sedentary Behavior (hours)	−.107	.239	.016	−.024
Device-Measured Drinking Intensity (g/dl)	−.031	−.123	.542	.450
Device-Measured Time-to-Peak TAC (min)	.027	−.076	.456	.483
Device-Measured Longest Duration of Transdermal Alcohol (min/day)	−.090	−.130	.320	.317

* p < .05, ** p < .01.

Drinking Intensity Models

Table 5 presents MLM coefficients from the over-dispersed Poisson models using self-reported (Model 1) and device-measured (Model 2) drinking intensity. For device-measured sedentary behavior minutes were rescaled to an hour scale. For device-measured physical activity step counts were rescaled to a 1,000-step scale. Contrary to hypotheses, both models indicated that the intensity of daily alcohol consumption was not associated with usual levels of or daily deviations in either physical activity or sedentary behavior. The social weekend was a significant covariate in both models; participants drank significantly more on social weekends (i.e., Thurs-Sat) than weekdays. Neither sex nor baseline BMI were significant covariates in either model.

Table 5.

Multilevel model coefficients examining associations between physical activity, sedentary behavior and drinking intensity

	Model 1: Self-Reported Drinking Intensity		Model 2: Monitored Drinking Intensity
	γ	SE	γ	SE
Intercept	−0.12	0.64	−4.11*	0.89
Between-person Physical Activity
Self-Report (MET-min)	0.26	0.32	–	–
Device-Measured (1,000 Step Counts)	–	–	−0.03	0.05
Within-person Physical Activity
Self-Report (MET-min)	−0.05	0.16	--	--
Device-Measured (1,000 Step Counts)	–	–	−0.02	0.01
Between-person Sedentary Behavior
Self-Report (Hours)	0.04	0.05	–	–
Device-Measured (Hours)	–	–	0.02	0.10
Within-person Sedentary Behavior
Self-Report (Hours)	−0.05	0.02	–	–
Device-Measured (Hours)	–	–	0.002	0.02
Sex	−0.26	0.24	0.10	0.31
Social Weekend	1.83*	0.18	1.14*	0.13
BMI	−0.02	0.02	0.002	0.03

* p < .01.

Timing of Alcohol Consumption Models

Table 6 presents MLM coefficients from the over-dispersed Poisson models of timing-based models of alcohol consumption. For device-measured sedentary behavior minutes were rescaled to hours. For device-measured physical activity step counts were rescaled to per every 1,000 step counts. Model 3 indicated that daily self-reported drinking duration was negatively associated with daily deviations in sedentary time but not usual levels of sedentary time. On days when participants reported an hour more than usual of daily sedentary behavior, they also reported drinking for less time than usual (γ = −0.06). Self-reported drinking duration was not associated with either usual levels of or daily deviations in physical activity. Self-reported drinking duration was longer on the social weekend than on weekdays, but was not associated with sex or BMI.

Table 6.

Multilevel model coefficients examining associations between physical activity, sedentary behavior and timing of daily alcohol consumption

	Model 3: Self-Reported Drinking Duration		Model 4: Monitored Time to Peak Drinking Intensity		Model 5: Monitored Longest Duration of Transdermal Alcohol Detection
	γ	SE	γ	SE	γ	SE
Intercept	−0.53	0.63	4.66**	0.49	5.74**	0.32
Between-person Physical Activity
Self-Report (MET-min)	0.04	0.32	–	–	–	–
Device-Measured (1,000 Step Counts)	–	–	−0.03	0.03	−0.03	0.02
Within-person Physical Activity
Self-Report (MET-min)	0.08	0.19	–	–	–	–
Device-Measured (1,000 Step Counts)	–	–	0.01	0.02	−0.03**	0.01
Between-person Sedentary Behavior
Self-Report (Hrs)	0.03	0.05	–	–	–	–
Device-Measured (Hrs)	–	–	0.02	0.06	−0.01	0.04
Within-person Sedentary Behavior
Self-Report (Hrs)	−0.06*	0.03	–	–	–	–
Device-Measured (Hrs)	–	–	−0.01	0.03	−0.01	0.02
Sex	−0.14	0.24	−0.48*	0.17	−0.12	0.12
Social Weekend	1.85**	0.20	0.73**	0.16	0.29**	0.09
BMI	−0.02	0.02	−0.01	0.02	−0.003	0.01

* p < .05, ** p < .01.

Model 4 indicated that daily device-measured time to peak drinking intensity was not associated with either usual levels of or daily deviations in physical activity or sedentary behavior. Time to peak drinking intensity was longer on social weekends than on weekdays (γ = 0.73) and was shorter for women than men (γ = −0.48) even after adjusting for BMI.

Model 5 indicated that daily device-measured duration of transdermal alcohol detection was negatively associated with device-measured daily deviations in physical activity but not usual levels of physical activity. On days when participants took 1,000 more steps than usual, the longest episode of continuous transdermal alcohol detection was shorter (γ = −0.03). Device-measured duration of transdermal alcohol detection was not associated with either usual levels of or daily deviations in sedentary behavior. Duration of transdermal alcohol detection was longer on a weekday than on a social weekend (γ = 0.29), but was not associated with either sex or BMI.

Discussion

This study examined whether between- or within-person changes in daily physical activity volume or sedentary behavior would be associated with drinking intensity or timing of daily alcohol consumption. Contrary to previous literature, this study found no association between daily physical activity and daily drinking intensity. This finding was consistent across self-reported and sensor data as well as for between- and within-person changes. It was not aligned with the majority of previous findings in which higher levels of physical activity were associated with higher levels of alcohol consumption [16–18]. One recent study with college students provided similar results; Henderson and colleagues did not detect significant between- or within-person associations [19]. Similar to the present study, they used a multi-method approach (i.e., self-reported alcohol use, accelerometer physical activity). It is possible that because previous research has largely relied on self-report measures, the positive association trend in the literature is inflated due to poor recall or social desirability bias.

Our unique sample might also explain this unexpected finding. The sample was comprised of high-risk college students who reported binge drinking recently and self-identified as polysubstance users within the last year. These high-risk individuals may not have the same incentives (e.g., cope with stress, conformity) for engaging in physical activity and alcohol use that could explain why these behaviors are positively associated in lower-risk populations of drinkers. This sample was also more active than normative college students. The majority of our sample reported meeting national aerobic physical activity guidelines (75%) compared to a little more than half (57%) in previous literature [4]. Capturing a positive association between physical activity and alcohol use may be more likely in populations who are not meeting the guidelines. Results may differ in samples that include not only those who are less active but also individuals who are less sedentary.

Results indicated that daily deviations in device-based physical activity (defined as step counts) were negatively associated with duration of transdermal alcohol detection. On days when participants took 1000 more steps than usual, the longest episode of continuous transdermal alcohol detection was shorter. This finding highlights a novel aspect of the topography of daily alcohol use (i.e., timing of drinking). Contrary to prior thinking, physical activity may actually be negatively associated with time-based measures of alcohol use. Causality cannot be inferred from this study; however, it is possible that the alcohol is metabolized more quickly on days when people accumulate greater physical activity, resulting in less time when alcohol is detected in the body. Research needs to uncover when in the drinking day these extra step counts occurred and if they occurred before, during, or after drinking episodes (e.g., taking extra step counts to walk between bars or to walk home).

This study also extended prior work on associations between alcohol use and sedentary behavior from adolescents to college students [3]. On days when participants reported an additional hour of sedentary behavior above their usual, they reported drinking for less time than usual. This finding refuted our hypothesis that sedentary behavior and alcohol use would be positively correlated. High-risk polysubstance using college students are faced with different obligations and social pressures compared to adolescents. One reason for the difference in these associations could be that adolescents might engage in sedentary behavior during their free time (i.e., watching TV or playing video games outside school hours) which may also be conducive for alcohol use if adolescents are unsupervised [51]. On the other hand, when college students engage in more sedentary behavior it could be due to academic responsibilities, which are less conducive for alcohol use. Our results are also specific to drinking duration whereas prior work focused on drinking intensity [3]. Similar to physical activity, the association between sedentary behavior and alcohol use may be negative when examining time-based measures of alcohol use such as drinking duration. Again, causality cannot be inferred from this study, but it is possible that additional sedentary time reduces leisure time outside of the home and motivation to go out, socialize and drink. On these days, drinking duration may be shortened because individuals spend less time in bars or nightclubs.

Limitations

Conclusions from this study may not generalize to those who are not high-risk drinkers and polysubstance users or those who have or are currently undergoing treatment for substance dependence. Variability in cannabis or cigarette use may influence associations between physical activity and alcohol use. Future research in this area should explore how varying levels of other drug use might impact these associations. This study design was observational and precluded causal inferences about associations between physical activity, sedentary behavior and alcohol use. Analyses focused on same-day associations using aggregated measures of daily behavior; conclusions may not generalize to lagged relations or research performed on other timescales.

Behaviors varied widely during the study (e.g., step counts ranged from <1,000 to >30 k steps/day, self-reported servings ranged from 0 to 10 servings/day, and drinking intensity ranged from 0 to 0.342 g/dl/day), but an 11-day monitoring period may not have been representative. Participants self-selected when they wanted to participate in the study and might have chosen a weekend with few obligations and more free time to engage in alcohol use. A selection bias of that nature could alter variability so future work may benefit from longer monitoring periods.

Conclusions

This study used intensive longitudinal methods to examine associations between physical activity, sedentary behavior and alcohol use among high-risk, polysubstance using college students. The antagonistic clustering of physical activity and alcohol use intensity identified in previous studies did not generalize to polysubstance-using college students engaged in high levels of both physical activity and sedentary behavior. Instead, physical activity and sedentary behavior were negatively associated with time-based measures of alcohol use (i.e. drinking duration and duration of transdermal alcohol detection). Future work can advance this discovery work by examining these associations on a momentary timescale and investigating the temporal sequencing of these behaviors.

Funding Sources:

This publication was made possible by a NIDA-funded predoctoral fellowship to ABW (T32 DA017629). Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIDA or NIH.

Compliance with Ethical Standards

Authors’ Statement of Conflict of Interest and Adherence to Ethical Standards Authors Ashley B. West, Rachel N. Bomysoad, Michael A. Russell, and David E. Conroy declare that they have no conflict of interest. All procedures, including the informed consent process, were conducted in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2000.

Authors’ Contributions

ABW: conception and design, acquisition of data, analysis and interpretation of data; drafting and revising; final approval of the version to be published. RNB: acquisition of data; critically revising; final approval of version to be published. MAR: conception and design, analysis and interpretation of data; critically revising; final approval of version to be published. DEC: conception and design, analysis and interpretation of data; drafting and critically revising; final approval of version to be published.

Ethical Approval All procedures performed in this study were in accordance with the ethical standards of our institutional research ethics committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Slight deviation from this concerned informed consent, as explained below, but this was approved by our ethics committee.

Informed Consent Consent was obtained from all individual participants included in the study. However, gives the nature of the research consent was not fully informed, although participants were aware that information was being withheld from them.